Part. The Methods of Political Science. Part

Transcription

1 Part The Methods of Political Science Part 1 introduced you to political science and research. As such, you read how to conduct systematic political research, decide on a potential topic, and conduct a systematic search for related literature so that you could write a comprehensive literature review. Part 2 introduced you to the scope of political science. We devoted several chapters to telling you about the political world, the American political process, public administration and public policy, comparative politics, and international relations. Now the relatively smooth road we traveled starts to get a bit bumpy. In Part 3, we discuss the basic elements of scientific research. We introduce you to terms such as concepts, variables, hypotheses, causation, and measurement. But do not worry. Our vehicle has brand new shocks, we drive slowly, and, as a result, we give you a smooth ride over the rough road we travel. In Chapter 9 we cover the elements of research. You will learn all about nominal and conceptual definitions, variables, and hypotheses. An understanding of Part 3 Chapter 9 Elements of Research: Determining Causal Relationships Chapter 10 Understanding Measurement Chapter 11 The Research Design Chapter 12 Data Collection and Input Chapter 13 Univariate Statistics Chapter 14 Bivariate Statistics Chapter 15 Multivariate Analysis Chapter 16 Putting It All Together

2 the chapter will enable you to identify the necessary factors to have a causal relationship and understand multiple causation. In Chapter 10 we introduce you to the notion of measurement, which is difficult for novice political researchers to grasp. We discuss the theory of measurement, the various threats to measurement, and how to enhance the reliability and validity of measurement instruments. We also explain how measurement is used to answer questions about topics such as voting turnout, the governmental structure of other nations, and why people revolt. Last, we discuss the different levels of measurement and their importance to the research process. In Chapter11 we discuss the different types of research designs you can use to complete a research study. Research designs provide plans you can use when collecting, analyzing, and interpreting data you believe provides answers to your research problem. As such, they are an important part of the research process. In Chapter 12 we deal with the various techniques political scientists use to collect and input data. We also examine the fundamentals of sampling theory and the aims of sampling. You will read about the central concepts of sampling, such as the population, the sampling unit, the sampling frame, and the sample itself. We also look at probability and nonprobability sampling designs. The latter part of the chapter covers ways to determine the size of the sample, coding of the data you collected, and inputting the data to a statistical analysis software package. In Chapters 13 through 15 we cover the various ways you can analyze your data. Chapter 13 deals with univariate statistics that are used to examine a single variable. In this chapter we look at some important tools to use in the preliminary stage of data analysis, such as frequency distributions; measures of central tendency, which describe the distribution s main characteristics; and measures of dispersion, which depict the extent of variance from the typical value or average. We also tell you how to use frequency polygons to determine how extreme scores can affect the measures of central tendency we discuss. The statistics we discuss in Chapter 13 help the political scientist understand data distributions. These descriptive statistics, however, are only a first step in data analysis. Once summarized, researchers often want to discover relationships between variables. We turn to this issue in Chapter 14 by discussing bivariate statistics, or the analysis of two variables. We introduce you to the rudiments of association and the various measures of association you can use to determine whether there is a statistical relationship between two variables. We also spend time discussing the important notions of statistical significance and hypothesis testing. Chapter 15 is a little more complex than previous chapters. In this chapter we talk about multivariate statistics, or the analysis of three or more variables. We cover control and ways to use statistics to determine causality and explain the political phenomena we observe. Cross-tabulation control procedures, partial correlation coefficients, and multiple regression techniques, for example, are useful because they allow us to examine the complex relationships among variables. In Chapter 16 we try to bring everything together by summarizing the steps in the research process and by examining and critiquing the efforts of other political science scope and methods students. We also talk about the possible theoretical and policy implications of your study. We conclude by suggesting ways to write an effective research report.

3 Chapter Chapter 9 Elements of Research: Determining Causal Relationships Outline 9-1 Introduction 9-2 Constructing Causal Explanations 9-2a Theory 9-2b Concepts 9-2c Variables 9-2d Hypotheses 9-2e The Population and Units of Analysis 9-2f Types of Data 9-2g Period of the Study 9-2h Causal Relationships 9-3 Multiple Causation Chapter Summary Chapter Quiz Suggested Readings Key Terms aggregate data antecedent variable causal relationship causation concept co-vary dependent variable hypothesis independent variable individual data intervening variable multiple causation negative relationship nominal definition operational definitions population positive relationship spurious relationship theory units of analysis variable 167

4 168 Chapter Introduction Having completed the literature review stage of the composite research process, you should be thoroughly familiar with existing writings, theory, and research in your area of interest. The next major stage is to identify and specify the various elements of your research project. As such, we will teach you about concepts, variables, hypotheses, and causation terms that are new to most of you. An understanding of this chapter will enable you to 1. Define theoretical population, units of analysis, concepts, variables, and hypotheses. 2. Identify types of data. 3. Identify the factors necessary to have a causal relationship. 4. Define multiple causation. 9-2 Constructing Causal Explanations In Chapter 1 we said that political scientists accept the notion that people act in a consistent manner. We also said that political science involves the study of people in order to explain their political behavior. These statements imply that the research process begins with an observation of behavior a process that looks for differences in the way people act and behave. We study political leaders, members of political parties and interest groups, and the citizens of America and other nations. As such, political scientists specify questions based on the observations they make. Typically they are interested in accounting for entire classes of events or relationships. Consider the following questions, for example. Why do some states spend more on social programs than other states? Why are some governments more authoritarian than others? Why do American voters belong to different political parties? As educated people, we assume that such variations are not just random. That is, states with larger budgets for social programs are different from those that spend less for such programs. It is your task to identify the differences that determine the success of social programs implemented in the states under study. Similarly, nations with authoritarian regimes differ from those without. Individual voters are different. Again, your task is to identify those differences among nations that cause them to differ with respect to government structure, or to explain differences among citizens that contribute to their affiliation with different political parties. In short, your task is to do research. Through the composite research process, you need to take those steps that will provide scientific explanations to questions similar to those we posed earlier in this section. In order to take the proper steps, however, we need to explain the building blocks of the research process. theory: An integrated set of plans intended to explain or account for a given phenomenon. 9-2a Theory A theory is an integrated set of plans intended to explain or account for a given phenomenon. The goal of theoretical research is to expand your knowledge about political events and why they occur as they do. Theoretical research is empirical and strives to discover facts about politics. It tries to develop new political theories, change existing theories, or confirm existing theories. As such, the major goal of theoretical research is to advance theories that link observed facts about politics (Shively 1990, 7). Anthony Downs, for example, developed what is called a spatial theory to explain how voters choose between political parties in American elections. The

5 Elements of Research: Determining Causal Relationships 169 basic argument of his spatial theory is that you can reduce all political issues to a single ideological scale and place all politicians and voters on the scale (we discuss scales in more detail in the next chapter). The scale can range from an extreme liberal position (0), to 100, which is an extreme conservative position. On this scale, you might place a liberal such as former Democratic presidential candidate George McGovern at 5. On the other hand, you might place a conservative such as former Republican presidential candidate Barry Goldwater at 85. To continue the example, you might place more moderate political figures such as former presidents Jimmy Carter and Gerald Ford toward the center of the scale, for example 30 and 70, respectively. Downs asserts that the voters know where the parties and their candidates stand on the issue scale. As a result, they vote for the party that best represents their beliefs and values. In America, those parties tending to have a centrist position have enjoyed the most electoral success. Thus, while individual politicians may find themselves somewhat off center on the scale, the two major political parties tend to seek a centrist position on the scale. The belief is that those parties that find themselves toward one pole or the other on the scale reduce their chances of electoral success. Therefore, if the parties major concern is to win elections, they should nominate candidates as close to the center as possible (Downs 1957). This may explain why political candidates are so eager to label their opponents as liberal or conservative and scoff when their opponents profess to have moderate political beliefs. It may also help explain why Goldwater and McGovern were soundly defeated at the polls. 9-2b Concepts We have repeatedly talked about the importance of problem definition. To accomplish this task successfully, you must define the problem in a way that will allow you to gather information about the problem. This step often requires you to conceptualize and operationalize. The problem must be broken down into concepts that allow measurement. We discuss measurement in the next chapter. Now, however, we turn our attention to the subject of political concepts and how you can make them subject to measurement through operationalization. A concept is an abstraction based on characteristics of perceived reality. Some examples in daily life include work, play, test, job, and motivation. Political scientists use concepts to represent political characteristics. They are the basic building blocks in answering political questions. Consequently, they are a beginning in constructing causal explanations. Concepts pinpoint an essential idea or element that you believe accounts for the entire subject of the study. Concepts summarize the critical aspects in a class of events. Examples of concepts used by political scientists are power, political efficacy, political socialization, freedom of expression, political ideology, political participation, and scope of government. concept: An abstraction based on the characteristics of some perceived reality. Political participation, for example, is a concept based on voting, campaigning, and running for office. Concept Operationalization Your ultimate goal is to scientifically analyze the concepts you use in your study. To do this, you must operationalize your concepts. Operational means measurable. Measurement and measurable have to do with numbers. Thus, you will need to assign numbers to the concepts you develop. You must determine how you can assign numbers to the concepts of power, political efficacy, political socialization, and political culture, for example. Simply put, you will take steps to translate the concepts you examine into observable and definable events by identifying indicators of your concepts. We call this important process concept operationalization.

6 170 Chapter 9 nominal definition: The dictionary definition of a concept. It is also referred to as a conceptual definition. operational definitions: The rules by which a concept is measured and scores are assigned. Nominal Definitions Concepts are the premises that provide an answer to your research question. To construct explanations of events or relationships you must define the concepts you intend to use in your study. There are two types of concept definitions: nominal definitions and operational or working definitions. A nominal definition is nothing more than the standard dictionary definition of the concept. For example, you may define patriotism as love for one s country. Many define intelligence as an individual s mental capacity. Or, occupation is often defined as the type of work one performs. As you can see from these definitions, you often use your discretion when developing nominal definitions. In order to advance knowledge, however, you should attempt to make definitions as realistic and sensible as possible. Thus, you should follow some general guidelines when developing a nominal definition. The definition must be clear enough that others can understand what you mean. If you adhere to the following suggestions, you will be able to satisfy this requirement. Ensure you have defined the concept rather than linking it to a related but different concept. Consider the following: Political interest is the extent to which one participates in the political process. In this example political interest is linked to political participation. The point is, however, that neither concept is defined. Make sure the definition is not circular. In other words, do not define the concept in terms of itself. For example: Political ideology means one s ideas about politics. It would be better to say Political ideology is a set of interrelated and consistent beliefs about the purpose, scope, and value pursuits of government (Janda et al. 2002). Operational Definitions Operational definitions translate the nominal definition into a form so we can measure the concept empirically. An operational definition is the working definition, or specification of the process by which a concept is measured. The operational definition specifies how we will measure a concept, so it should be clear and complete. Two examples that come to mind are age and intelligence. We can operationally define age as the number of years the person, object, or law has existed. To operationalize intelligence, we might use criteria such as grade point averages and IQ scores. Operational definitions allow you to assess the extent of empirical support for your theoretical ideas. To clarify our point, recall the first research question we posed in Section 9-2, Constructing Causal Explanations : Why do some states spend more on social programs than other states? As we said before you must assume that such variations are not just random. States that spend more on social programs are different from those with small social program budgets. To resolve this research question you need to identify differences in the social, political, and economic activities in the states. Then you must collect data that, with analysis, will suggest possible explanations. The concepts you choose to answer this question could include fiscal federalism, political socialization, political culture, and economic development. To operationalize these concepts, you would collect data concerning levels of federal and state funding, the levels of education and income in the state, the geographical location of the state, and the level of unemployment in the state. Before we summarize this section, let s look at nominal and operational definitions of political party identification.

7 Elements of Research: Determining Causal Relationships 171 Nominal definition: Political party identification is nominally defined as affiliation with a political group organized to promote and support its principles and candidates for public office (Webster s II, 1999). Operational definition: Political party identification is operationally defined in terms of responses to the following question from the 2000 NORC GSS. Generally speaking, do you usually think of yourself as a Republican, Democrat, Independent, or what? 1) Republican 2) Democrat 3) Independent 4) Other In sum, you continue the research process by defining your research problem in terms of nominal and operational concepts. Then you must find ways to collect data that represents the concepts you use to answer your research question. It is not an easy process. Most beginning research students have more difficulty conceptualizing theory and operationalizing concepts than any other stage of the research process. It takes practice and familiarity with literature on your subject for you to sharpen your skills. 9-2c Variables Variables are concepts that you operationalize, or measure, in a sample of data. In short, a variable is a measured concept. You use variables to move from the conceptual level to the empirical level of research. Variables assign numerical scores or category labels to each item in your sample. If you want to study the extent of political participation of college students, for example, you might gather data for one hundred students on the variables of gender, age, grade level, grade point average, major, income, political party identification, voting record, and political campaign activity. There are two major types of variables: dependent and independent. The dependent variable is the variable you are trying to explain. It is the effect variable. For example, voter turnout, political party identification, and political tolerance are variables that change because of the impact of other variables. The independent variable is the variable that contributes to the explanation of the dependent variable. It is the causative variable, or the predictor variable. We use the independent variables to predict the dependent variable. Common examples include voter registration, attitude toward a public policy, and church membership. You use these variables to examine their effect on the variable you want to explain. Consider, for example, the following diagram relationships between the variables discussed in this section. Independent Variable Dependent Variable Registration difficulty leads to voter turnout. Support for legal abortion leads to political party identification. Church membership leads to political tolerance. In the table s examples, registration difficulty is used to explain voting turnout, support for legal abortion is used to explain one s political party identification, and religious affiliation is used to explain one s political tolerance. The primary purpose of political research is to determine the effect that an independent variable has on a dependent variable. There are, however, types of variables that could affect the relationship you find between an independent variable and a dependent variable. We call these variables antecedent variables and intervening variables. variable: A measured concept. dependent variable: The phenomenon thought to be influenced, affected, or caused by some other phenomenon. independent variable: The phenomenon thought to influence, affect, or cause some other phenomenon.

8 172 Chapter 9 antecedent variable: An independent variable that precedes other independent variables in time. intervening variable: A variable that occurs in time between an independent and dependent variable in an explanatory scheme. An antecedent variable is an independent variable that precedes other indepent variables in a time. An antecedent variable could affect the independent variable and alter its relationship to the dependent variable. Consider this example. You want to identify those variables that explain political party identification (PID). As a result of your literature review and data analysis, you decide that one s attitude about legal abortion helps to explain why your subjects affiliate with the Republican or Democratic Party. In this example, PID is the dependent variable, and attitude about legal abortion is the independent variable. That is, attitude helps to explain PID. You may, however, also want to know what explains your subject s attitudes about legal abortion. So you decide to examine the effect of attendance at parochial schools. Several possibilities about the original relationship can arise when you examine the effect. First, you do not see a difference in the original relationship. Thus, the relationship remains as originally diagramed. Second, you could find that it changes the original relationship between support for legal abortion and one s political party affiliation. That is, you find that individuals who attended parochial schools have a different attitude about abortion from those who attended public schools. Thus, the antecedent variable affects one s support for legal abortion while altering its relationship with political party identification. Hence, we can diagram the relationship like so: Antecedent Original Original Variable Independent Variable Dependent Variable Type of school leads to support for legal leads to political party attended abortion identification. It is also possible that the antecedent variable affects both the independent variable and the dependent variable, with the original relationship virtually disappearing. This is a spurious relationship, which we will discuss in Section 9-2h. This type of relationship can be diagramed like so: Type of school leads to support for legal attended abortion. Type of school leads to political party. attended identification An intervening variable occurs between an independent and dependent variable and affects the relationship between them. Several political researchers, for example, have used education and type of school attended as independent variables to explain different political concepts. It is doubtful, however, that the years someone spent in school had a direct effect on their choice of political party. It is more likely that experiences in school lead to other qualities that affect one s party affiliation. Exposure to socially and economically diverse groups could lead to affiliation with a political party. We can diagram the connections between these variables as follows: Type of school leads to club membership leads to political party attended identification. The type of school attended (independent variable) affects exposure to social diversity (intervening variable), which in turn affects political party identification (dependent variable). The identification of antecedent and intervening variables allows you to develop a more complete understanding about the political question you want to resolve.

9 Elements of Research: Determining Causal Relationships d Hypotheses Hypotheses are statements that formally propose an expected relationship between independent variables and dependent variables. The key word in this definition of a hypothesis is expected. This is a key word because hypotheses are tentative answers to research problems tentative because you cannot verify a hypothesis until you test it empirically. The components of a hypothesis are a dependent variable, at least one independent variable, and an expected relationship between the variables. Consider the following hypothesis using our example involving attitude toward legal abortion (the independent variable) and political party identification (the dependent variable): Individuals who support legal abortion will identify with the Democratic Party, while individuals who do not support legal abortion will identify with the Republican Party. hypothesis: A statement proposing an expected relationship between two or more variables. The Purpose of Hypotheses The primary purpose of hypotheses is to account for changes in the dependent variable by linking criteria between independent and dependent variables. In the example we presented earlier in this section, we hypothesized that variation in political party identification is due to one s attitude about legal abortion. Notice that our hypothesis reflects the expected relationship by linking supporters with the Democratic Party and nonsupporters with the Republican Party. The primary virtue of hypotheses is that they allow theoretical ideas and explanations to be tested against actual data. Thus, they provide a means to evaluate theory. Requirements of Sound Hypotheses In this section we discuss the requirements of sound hypotheses and offer examples that do not meet the requirements. We also show you how to reword each hypothesis so that it fulfills the requirements we are about to discuss. There are several requirements of sound hypotheses. First, the concepts must allow measurement and empirical testing. In other words, you must be able to operationalize them. Remember your goal is to evaluate theory and provide explanations for your research question through the empirical testing of the data you collect. Second, you must state hypotheses clearly. That is, you must state in precise language the relationship between the independent and dependent variables. For example, what change in the dependent variable can you expect if there is a change in the independent variable? It is not sufficient to state that age affects voting. A proper hypothesis would read this way: Older people tend to vote Democratic. Third, you need to make hypotheses specific. In your hypotheses you need to assert that specific variation in the independent variable results in specific variation in the dependent variable. Consider: The greater the concentration of minorities in a community, the greater the demand for social programs. In short, you need to succinctly state how the variables in your hypothesis are related. Fourth, hypotheses are value free. In other words, they must be nonnormative. It is difficult, if not impossible, to resolve questions of good and evil through empirical research. Consider: Liberals are better than conservatives. While one might agree with the statement, it is value laden and cannot be empirically tested. On the other hand, the following statement can be empirically tested: Liberals are more supportive of programs that promote equality than conservatives. To summarize our discussion and illustrate the requirements we discussed, consider the following two examples of sound hypotheses. The greater the

10 174 Chapter 9 positive relationship: A relationship in which the values of one variable increase (or decrease) as the values of another variable increase (or decrease). A positive relationship is often referred to as a direct relationship. negative relationship: A relationship in which the values of one variable increase (or decrease) as the values of another variable decrease (or increase). education level, the greater the level of voter turnout. The lower the federal funding level for a program, the less chance the program will be successful. Each statement contains independent and dependent variables. They also depict a relationship between the variables. These are examples of a positive relationship between the variables. They show how increases or decreases in the independent variable are expected to lead to increases or decreases in the dependent variable. Some hypotheses, however, can be stated as a negative relationship in that an increase or decrease in the independent variable will lead to a decrease, or increase, in the dependent variable. For example: The higher the degree of federal restrictions on administering a program at the local level, the less is the chance of program success. Positive and negative relationships between variables refer to the direction of a relationship. Common Errors When Stating Hypotheses Students engaged in their first attempt at research often make several errors when developing hypotheses. Robert A. Bernstein and James A. Dyer offer some suggestions you should consider when writing hypotheses (Bernstein and Dyer 1992, 10 12). Following are their suggestions accompanied by our examples. Hypotheses Must Relate Two Variables Many students develop hypotheses having only one variable in the statement. Consider this example: Americans are not politically active. There is only one variable in this hypothesis, political activity. You need to relate political activity to another variable to have a sound hypothesis. Perhaps you should state the hypothesis this way: Americans with a low level of political efficacy are less likely to vote than those with a high level of political efficacy. The Relationship Is Unclear You must clearly state the relationship you expect to find between the two variables in your hypotheses. Here is an example of a vague hypothesis: Political efficacy is related to voting turnout. We cannot tell from this statement whether voting turnout increases or decreases as political efficacy increases or decreases. In sum, we do not know the direction of the relationship in the example. A sound hypothesis is As the level of political efficacy increases in a community, the level of voter turnout tends to increase. The Statement Lacks Generality Do not personalize your hypotheses. That is, they should not contain the names of individuals or countries. In addition, do not limit your hypotheses to a specific period of time. The following are examples of hypotheses that lack generality. The German government is more stable than the government of Russia. This statement is too specific. You are merely comparing the governments of Germany and Russia. A more general hypothesis is Governments based on long-standing democratic and capitalist premises are more stable than former communist regimes. Another example is The federal government distributed more funds to the states in 1966 than in Again, the hypothesis should be more general. A better statement is The federal government distributes more funds to the states during Democratic administrations than during Republican administrations. The Hypothesis Makes a Value Judgment Avoid hypotheses that make value statements. Remember, a characteristic of scientific research is the fact that it is not normative. Thus, do not use words such as

11 Elements of Research: Determining Causal Relationships 175 should, ought, better, or worse. Hypotheses with these words indicate that you made a value judgment when you wrote the hypothesis. For example, the following is not a sound hypothesis: All governments in the world should practice democracy. While many of us would prefer such a state of affairs, this hypothesis clearly depicts the values of the researcher. How would you collect data to analyze such a statement? Even with data, are you really proving the statement? If you examine the hypothesis, however, you may see you can write a sound hypothesis from this statement. For example: Democratic countries have higher levels of economic development than those countries without democratic governments. In sum, hypotheses establish a relationship between two or more variables. In doing so they demonstrate that there is an association between the independent variable and the dependent variable. You use hypotheses to demonstrate that the results are generally true in the real world. They also enhance your research effort because they reveal whether one phenomenon precedes another in time, and they eliminate as many alternative explanations for a phenomenon as possible. 9-2e The Population and Units of Analysis The theoretical population is the group of objects you study in order to produce answers or explanations to your research question. To clarify our discussion, let s return to the political questions we introduced in Section 9-2, Constructing Causal Explanations. Why do some states spend more on social programs than other states? Why are some governments more authoritarian than others? Why do American voters belong to different political parties? In these questions, the theoretical populations include American states, the nations of the world, and the American electorate. Within each theoretical population, you will collect data for individual states, nations, and voters to determine whether your explanation is correct. We refer to these individual objects within the population as units of analysis. When you collect data about social programs in states, as in the first question we just posed, each state is a unit of analysis. California might be one such unit. Thus, you would collect data about California that might explain the level of success for the state s social programs. Texas would be a second unit of analysis. For the question dealing with authoritarian nations you would use the same procedure. Iraq, Germany, Iran, Sweden, and Cuba might be the units of analysis you would analyze to explain why some nations are more authoritarian than others. To answer our question dealing with American voters and political party membership, you might gather data about Jerry Perry, Bill Henderson, and Robert Garza, who, as individual voters, make up part of the theoretical population you would analyze to determine why American voters belong to different political parties. A key to a successful scientific analysis, however, is to collect the same data for each unit of analysis. For each state in your analysis you would collect data about poverty, for example. For each nation of the world you are examining, you might collect data about its gross domestic product. And for each voter in your study, you would probably collect data about their race or religious affiliation. A final comment about populations and units of analysis concerns the scientific goal of generalization. There is a relationship between the range of an explanation and its value to the field of study. An explanation that applies to more people for a greater span of time has more value than one that applies to fewer people for a more limited time. population: All of the cases or observations covered by a hypothesis; all the units of analysis to which a hypothesis applies. units of analysis: The type of element (individual, group, institution, state, nation) that is specified in a researcher s hypothesis.

12 176 Chapter 9 individual data: Data that describes a single unit of analysis. Data, for example, that describes a single student or a member of Congress. aggregate data: Data that is based on groups of units that are put together, such as data for states, cities, or nations. Aggregate data is often referred to as ecological data. 9-2f Types of Data Different research efforts involve the analysis of different types of units of analysis. In addition, the different types of units of analysis are distinguished by two basic types of data: individual data and aggregate data. Individual data is data that is based on single units rather than on a collection of units. Data collected in surveys is the best example of individual data. In surveys, the units of analyses are people or individuals. When you analyze congressional voting preferences, the units of analyses are congressional members and the data is individual data. Aggregate data are often referred to as ecological data. Aggregate data is data that is based on groups of units put together. Census data, for example, that is collected and put together is a good example of aggregate data. When you analyze cities, states, or nations of the world, the data is presented for your analysis in the aggregate. For each of these units we study data in the form of averages, ratios, sums, and rates that describe the units of analyses. 9-2g Period of the Study Determining the period of the study seems like a relatively simple task. There are some questions, however, that you need to consider before deciding on the period of your study. For example, what are the time limits for your study? One year? Five years? Ten years? Is the period long enough to reflect a trend that, with analysis, answers your research question? Is the data available for the period you decide to examine? Thus, maybe this task is not so easy after all. Our suggestion is to decide on a time period to study that will produce reliable and valid data that answers your research question. If you are examining changes in public opinion, for example, you should probably examine survey results for several decades. After all, time is an important factor influencing people s attitudes. What was once tolerated within areas of our nation may now be improper or immoral. Conversely, what was once renounced by society may now be accepted by much of society. causal relationship: A relationship that occurs when the variation in one variable independent of variation in other variables causes variation in a second variable. causation: A phenomenon that involves three distinct operations: establishing the time order of the occurrences, demonstrating co-variation, and eliminating spurious relations. Causation also requires theoretical justification. 9-2h Causal Relationships A relationship, or correlation, in research broadly refers to any relationship between two or more variables. A causal relationship is a relationship between variables that occurs when changes in one variable are systematically related to changes in another variable. This is the type of relationship political scientists want to discover. A relationship between variables, however, does not necessarily mean that a causal relationship exists. Remember, correlation does not necessarily mean, or guarantee, causation. In other words, the observed relationship is merely a coincidence. As a reminder, direction of a relationships refers to positive or negative relations between variables. A positive relation means that as values of one variable increase, or decrease, values of the other variable also increase, or decrease. A negative relationship means that as values of one variable increase, or decrease, values of the other variable change in the opposite direction. The magnitude of a relationship between variables is also important when considering causality. The magnitude of a relationship is the extent to which variables change together in one direction or the other. The highest magnitude of relationship is a perfect relationship, in which knowledge of the value of the independent variable determines the exact value of the dependent variable. On the

13 Elements of Research: Determining Causal Relationships 177 other hand, the lowest magnitude of relationship, the zero relationship, occurs when systematic change between the values of an independent variable and a dependent variable is not discernable. There are four formal criteria that are necessary to establish that a relationship is causative: time order, covariance, nonspuriousness, and theoretical justification. Time Order A relationship is causal if an action is the result of another action. If A is the cause of B, then A must precede B in time. That is, changes in an independent variable (A) must occur before changes in a dependent variable (B). This is another way of saying that cause must precede effect. The notion of time order is probably the most intuitive of the criteria to demonstrate causality. Most times, it is relatively simple to determine time order. For example, gender and race precede political socialization and other political phenomena. In fact, we challenge you to identify some phenomena that precede these individual characteristics. At times, however, time order is more difficult to determine, for example, job tenure and job satisfaction. Does tenure lead to satisfaction? Or does satisfaction lead to job tenure? Co-Variation When we say that two variables co-vary, we mean that the two variables change, or fluctuate, together. That is, if the independent variable changes and the dependent variable also changes, the independent variable may be the cause of the dependent variable. In other words, if B changes when A changes, there may be a relationship because the two co-vary. However, if changes in A are never accompanied by changes in B, there is no co-variation and A cannot be the cause of B. Consider this hypothesis: Increases in salary lead to increases in productivity. If salary increases did lead to increases (or even decreases, for that matter) in productivity, then we may have support for the hypothesis. If, on the other hand, salary increases did not lead to changes in productivity, we can reject the hypothesis because the variables did not change together. Nonspuriousness This criterion of causal relationships is, perhaps, the most difficult to understand. A nonspurious relationship between two variables is an association, or covariation, that you cannot explain with a third variable. A spurious relationship results when the impact of a third variable explains the effect on both the independent and dependent variables under analysis. In addition, the original relationship between two variables virtually disappears when you consider the effect of a third variable. Consider this example. In her study about the causes of juvenile delinquency, a novice researcher thought she found a causal relationship between ice cream sales and juvenile crime. Those areas of the community having a high level of ice cream sales also had a high level of juvenile crime. She also discovered that the criteria of time order and co-variance applied to her findings. But does a third factor explain both the level of ice cream sales and juvenile crime? In other words, was this a nonspurious relationship? Further research indicated that those areas with a high rate of juvenile crime also had more juveniles living in the neighborhood. Thus, a younger population contributed to both increased ice cream sales and increased levels of juvenile crime. In short, a third variable explained the change in both the independent and dependent variables. co-vary: The situation where a unit change in one variable (i.e., education) is paralleled with some degree of regularity by a comparable change in another variable (i.e., income). spurious relationship: A relationship between two variables that is caused entirely by the impact of an antecedent variable.

14 178 Chapter 9 In summary, you will find it difficult to demonstrate the criterion of nonspuriousness. To accomplish this task you must consider all possible variables and their effect on the independent and dependent variables. In statistical analysis we say that you must control for the effects of a third variable. If you find that change in a third variable explains change in the independent and dependent variables and the original relationship between the variables virtually disappears, the original relationship is spurious. Conversely, if you find that a third variable does not affect the independent or dependent variable, then the original relationship is nonspurious. As such, you will have satisfied an important criterion of causality. We discuss the procedures you take to control for the effect of a third variable in Chapter 15. Theoretical Justification It is unlikely, but possible, that you will satisfy the first three criteria for causality and still not have causality. That is, your independent variable may precede the dependent variable in time, the two variables may vary together, and you may not find a third variable that explains change in the variables and the original relationship. It may just so happen, however, that the relationship does not make sense. Consider our example about ice cream sales and juvenile crime. Does it make sense that ice cream sales cause incidents of juvenile crime? Thus, not every association is a causal one. Demonstrating that there is an association between independent and dependent variables is not sufficient to verify a hypothesis. You need to find a causal association in which variation in one variable influences variation in another variable. In some cases it may be difficult for you to ascertain whether a relationship makes sense. Thus, you need to make sure that your independent variables have theoretical justification. A theoretical or substantive justification for the relationship must be provided to support a relationship between variables. A theory interprets the observed co-variation while addressing the issue of how and why the relationship occurs. A theory can also serve as an additional check to ensure the relationship you discovered is not spurious. In summary, theory provides support and a substantive justification for observed relationships between two variables. multiple causation: An instance when a dependent variable is affected by two or more independent variables. 9-3 Multiple Causation Most times a single variable will not totally account for change in another variable. We live in a complex and changing environment. Rarely can we explain the behavior of humans and their political institutions through the analysis of a single factor. Similarly, a single hypothesis is not apt to provide a complete explanation. That is, there are several independent variables that explain why some dependent variable changes. This is known as multiple causation. Age, income, education, race, registration requirements, the type of election, and other variables may influence voter turnouts. Finding a complete explanation, however, is almost impossible. As scholars, we try to add to existing explanations in incremental steps. That is, many research efforts have limits imposed by time and other resources. Therefore, most research results in incomplete explanations. According to Bernstein and Dyer, however, research... is the accumulation of those incomplete explanations that advances our knowledge of human behavior (Bernstein and Dyer 1992, 15).

15 Elements of Research: Determining Causal Relationships 179 Summary Chapter Summary In this chapter we discussed ways to identify and specify the various elements of your research project. As such, we talked about concepts, variables, hypotheses, and causation. In the next chapter we introduce you to the idea of measurement, or the assignment of numbers to some phenomenon you are interested in analyzing. Quiz Chapter Quiz 1. developed what is called a spatial theory to explain how voters choose between political parties in American elections. a. Robert Dahl b. Anthony Downs c. David Easton d. Harold Lasswell 2. A research project calculated the percentages of students who are female at a sample of colleges and universities. The unit of analysis in this study was a. females. b. gender. c. students. d. universities and colleges. 3. Which of the following is not a properly stated hypothesis? a. There is a positive relationship between support for gun control and race. b. Men are more supportive of sexual equality than women are. c. There is no relationship between income and education. d. There is an inverse relationship between a state s unemployment rate and a state s level of education. 4. Antecedent variables are always a. important for explaining a relationship between variables. b. causally prior to the independent variable. c. interactive variables. d. spurious. 5. Suppose you are studying the effects of income on people. Which of the following could not be used as a dependent variable in your study? a. race b. views on capital punishment c. views on welfare programs d. political tolerance 6. A(n) is a statement or series of statements that organize, explain, and predict knowledge. a. research plan b. theory c. literature review d. concept 7. A(n) is an abstraction based on characteristics of perceived reality. a. concept b. variable c. theory d. index 8. are concepts that researchers operationalize, or measure, in a sample of data. a. Antecedent variables b. Intervening variables c. Dependent variables d. Independent variables e. All of choices a through d are operationalized concepts. 9. Ecological data is often referred to as data. a. survey b. individual c. aggregate d. primary 10. A(n) is a variable that occurs, in time, before the other variables being studied in a research project. a. dependent b. independent c. antecedent d. intervening

16 180 Chapter 9 Readings Suggested Readings Bernstein, Robert A. and James A. Dyer. An Introduction to Political Science Methods, 3rd ed. Englewood Cliffs, NJ: Prentice-Hall, Downs, Anthony. An Economic Theory of Democracy. New York: Harper and Row, Fox, William. Social Statistics, 3rd ed. Bellvue, WA: Micro- Case, Frankfort-Nachmias, Chava and David Nachmias. Research Methods in the Social Sciences, 6th ed. New York, NY: Worth Publishers, Goldenberg, Sheldon. Thinking Methodologically.New York: HarperCollins, Janda, Kenneth, Jeffrey Berry, and Jerry Goldman. The Challenge of Democracy, 7th ed. Boston-New York: Houghton Mifflin Company, Johnson, Janet Buttolph, Richard A. Joslyn, and H. T. Reynolds. Political Science Research Methods, 4th ed. Washington, D.C.: Congressional Quarterly Press, Kay, Susan Ann. Introduction to the Analysis of Political Data. Englewood Cliffs, NJ: Prentice-Hall, 1991.